Performances analysis of parallel asynchronous and synchronous algorithms on cloud architecture for PDE

Abstract

In order to solve discretized stationary linear or non linear diffusion and convection–diffusion problems, the present paper deals with a comparison between the performances of parallel iterative synchronous or asynchronous algorithms implemented on a Cloud computing architecture with several sizes of the algebraic systems to be solved. The parallel algorithms are briefly and intuitively presented and analysis of their behavior is recalled. In particular, criteria of convergence are recalled and estimates of convergence speed are given. Implementation on cloud architectures are briefly described. Finally, for each kind of problem, the performance of the target algorithms on cloud architectures are analyzed.